Instrument · the thesis estimator's own output

What the estimator believed.

Every minute, the frozen model holds a probability for each headcount from 0 to 160. Explore what it concluded on each exported day: all days at once, one day in depth beside the four radio signals it read, or several days overlaid. Every chart says what it shows and how to read it.

Frozen model 964a0a420ed0e427… · 16 exported days · exported 2026-10-17 15:57 UTC. Posteriors are filtered: each minute is answered from earlier minutes only and never revised. No people count is shown here, and none enters the model. How the estimator works, lesson by lesson.

Bands: 0 · 1–10 · 11–30 · 31–60 · 61–100 · 101+. Levels: empty = 0, low = 1–30, busy = 31+. Both declared scale priors are available; prior A is primary. "dev" days lie inside the frozen model's fit window, so they show how it reads observations, not how well it counts. JSON: the index · this day.